Image Restoration Using Dynamic Wavelet Thresholding and Artificial Neural Network

Images are corrupted by various means during its acquisition, processing, compression, transmission and reproduction. A host of techniques are available which have explored ways and means to improve the quality of restoration. This paper presents an image restoration approach performed by de-noising an image using Discrete Wavelet Transform (DWT) and Artificial Neural Network (ANN). The process assists in recovering the degraded image first by adapting a dynamic DWT and secondly by ANN to minimize the error to an extent which help in achieving satisfactory quality suitable for applications. The results obtained from both the process are compared. Both offer certain advantages and disadvantages and can be combined together to formulate a reliable image restoration technique. In this paper a set of experiments are performed using a range of images. The individual results obtained through both dynamic DWT and ANN shows a significant increase in percentage gain of peak signal to noise ratio (PSNR) reducing the effect of added noise in the image. The PSNR values calculated through ANN approach shows the better results than the dynamic DWT. The ANN based approach provides 6 to 26% improvement for a range of images considered.